Pacific bidecadal climate variability regulated by tidal mixing around the Kuril Islands



[1] 18.6-year period variability has been detected in various aspects of the climate, especially in and around the Pacific Ocean. Although it is believed to be caused by 18.6-year period tidal cycle, no study has directly shown how the tidal cycle regulates such variability. Using a state-of-the-art climate model, we show that the 18.6-year cycle in strong tidal mixing localized around the Kuril Islands induces 18.6-year periodicity in El Nino-Southern Oscillation-like Pacific climate variability. Influence of the tidal mixing propagates along the Pacific western rim as coastally trapped waves. Temperature anomaly is generated in the subsurface western equatorial Pacific, which propagates along the equatorial thermocline and eventually excites 18.6-year periodicity in the equatorial sea surface temperature.

1. Introduction

[2] The climate over the Pacific is known to have several distinct modes of variation from interannual to interdecadal time scales, among which the mode related to El-Nino and Southern Oscillation (ENSO) accounts for the largest part of sea surface temperature (SST) variability (Figure 1a). ENSO also affects climate of regions remote from the Pacific due to atmospheric teleconnection [Alexander et al., 2002], making it one of the primary concerns of global climate variability and predictability. At low frequencies, there is a mode known as ENSO-like decadal (or interdecadal) variability or Pacific Decadal Oscillation (PDO) [Mantua et al., 1997; Deser et al., 2004]. Its spatial structure resembles that of ENSO, but it exhibits dominant variability on interdecadal timescales [Minobe, 1999].

Figure 1.

First EOF mode of Pacific annual-mean SST north of 20°S. (a) Observed data for 1950–1999 [Rayner et al., 2003], (b) the case CONST, and (c) the case VAR. The observed mode corresponds well to the SST anomaly pattern of El Nino years, and its time series coincides well with widely-used ENSO indices [Wallace et al., 1998]. Unit is °C per standard deviation of the corresponding time series. Percentage variance of the mode against the total variance is indicated at the top of each panel.

[3] Bidecadal signals in the ocean and atmosphere in the Pacific are increasingly thought to be of lunar origin [e.g., Loder and Garrett, 1978; Currie and O'Brien, 1988; Osafune and Yasuda, 2006; Wilson et al., 2007]. The moon's orbital surface is inclined by 23.4° to the earth's equatorial surface on average, and this inclination oscillates with a period of 18.613 years and amplitude of 5.2° [Bradley, 1748]. Link has been discussed between this 18.6-year period nodal tidal cycle (18.6-year tides, henceforth) and various modes of Pacific climate variability, such as ENSO [Cerveny and Shaffer, 2001], PDO [Yasuda et al., 2006], and PNA [McKinnell and Crawford, 2007]. However, its physical mechanism has not been clarified.

[4] Tides affect mean state and circulation of the ocean primarily through their interaction with topography, manifested as tidally rectified flows [Young, 1983] or mixing induced by internal tides [St. Laurent and Garrett, 2002]. Straits along the Kuril Islands are known to be a location of strong vertical mixing forced by diurnal tides [Reid, 1965], and the resulting diapycnal meridional overturning circulation is thought to regulate the formation of North Pacific Intermediate Water (NPIW) [Tally, 1993; Yasuda, 1997] and associated circulation in the Pacific intermediate depths [Nakamura et al., 2004; Tatebe and Yasuda, 2004]. Representation of NPIW in global scale ocean models is reported to be improved by prescribing higher-than-usual vertical mixing around the Kuril Islands. An analysis of transient response to sudden rise of the vertical mixing shows that the effect of changes of mixing around the Kuril Islands propagates over the Pacific by coastal and planetary waves and modifies the basin scale mid-depth circulation [Nakamura et al., 2004]. Since the amplitude of diurnal tides is modulated by up to 20% by the 18.6-year tides, the change of the tidally induced mixing around the Kuril Islands could significantly influence the climate over the Pacific through the above-mentioned oceanic processes, as previously hypothesized [Yasuda et al., 2006].

2. Model and Experimental Design

[5] A state-of-the-art numerical climate model is utilized to elucidate the influence of the 18.6-year tides on climate. The model used here is the medium resolution version of CCSR/NIES/FRCGC MIROC 3.2 [Hasumi and Emori, 2004], which calculates states and interactions of various elements of the climate system (atmosphere, ocean, land and cryosphere) by specifying solar radiation and atmospheric concentration of radiatively active gases. The atmospheric component has 2.8° horizontal resolution and 20 vertical levels, and the oceanic component has 1.4° zonal resolution, 0.56°–1.4° meridional resolution (enhanced around the equator), and 44 vertical levels. This model has been shown to reasonably simulate the twentieth century climate [Nozawa et al., 2005] and climate variability from interannual to decadal time scales [Shiogama et al., 2005; Oka et al., 2006]. Vertical diffusivity in the ocean is estimated by a turbulence closure scheme which solves generation, advection, and dissipation of turbulent kinetic energy with its input from wind-waves parameterized [Noh and Kim, 1999]. For depths below the surface mixed layer, however, calculated turbulence is very small, and prefixed background values are mostly applied. The background values depend only on depth, taking 1 × 10−5 m2 s−1 for the surface and 3 × 10−4 m2 s−1 for the bottom-most model level (5500 m) with a sharp change around 1500 m. This background profile is an empirical one to reasonably reproduce the Pacific deep overturning circulation [Tsujino et al., 2000], and is applied to the entire ocean for the control climate calculation (see below), except for low latitudes where the values are reduced as a function of latitude [Gregg et al., 2003].

[6] The climate model is first integrated for 600 years by fixing the external forcing (in terms of solar constant, atmospheric composition, and aerosols) at the year 1850 (pre-industrial condition) to obtain a steady control climate. The background ocean vertical diffusivity along the Kuril Islands (over a two-grids-wide zone from the tip of Kamchatka Peninsula to Hokkaido) is then raised to 2 × 10−2 m2 s−1 over the full depth, as suggested by a direct calculation of tidal effects [Nakamura et al., 2000], and the calculation is continued for 200 more years (case CONST). Another experiment is initiated from the 100th year of the case CONST, where the background vertical diffusivity along the Kuril Islands is sinusoidally varied with an amplitude of 3 × 10−3 m2 s−1 and a period of 18.6 years (case VAR), based on the fact that the amplitudes of two major diurnal tide components, K1 and O1, vary by 11% and 19%, respectively, at 45°N [Loder and Garrett, 1978]. This calculation is also continued for 200 years. Since it takes a few decades until the North Pacific intermediate depths are adjusted to the switch of mixing at the Kuril Islands, the last 150 years are used for analysis in each experiment.

3. Results

[7] In both model calculations the first mode of empirical orthogonal function (EOF) analysis for the Pacific SST captures basic features of ENSO-like Pacific climate variability (Figure 1) as discussed in a previous analysis of the same model [Shiogama et al., 2005], although the equatorial high signal region extends westward compared with the observation as in other coupled climate models [e.g., Deser et al., 2006; Wittenberg et al., 2006]. The corresponding time series exhibits a distinct spectral peak centered at around 18.6 years in the case VAR, but not in the case CONST (Figure 2a). The peak in the case VAR exceeds the 90% confidence limit relative to the corresponding red noise spectrum. Furthermore, there is no overlap between the 90% confidence intervals of the two spectra at around 18.6 years (figure not shown), which means that the two spectra are statistically different. Therefore, only the case VAR exhibits a statistically significant spectral peak at around 18.6 years. Correlation between this time series and the 18.6-year tidal cycle becomes maximum (0.35, which exceeds 95% significance) at a 6-year lag with the tidal phase leading, and the lowest tropical SST is realized about 6 years after the strongest tidal mixing (Figure 2b). 6-year-lagged regression of annual-mean SST of the case VAR to the tidal cycle exhibits a region of statistically significant (more than 95%) correlation whose spatial pattern coincides with the strong signal region of the first EOF mode (figure not shown). Thus, the 18.6-year periodicity of the ENSO-like variability is a robust feature in the case VAR.

Figure 2.

(a) Thick line: power spectrum of the normalized (such that the standard deviation becomes unity) time series of the first EOF mode for the cases CONST (blue) and VAR (red). Thin line: 90% upper confidence limit against the corresponding red noise spectrum. (b) The 7-year running-averaged normalized time series of the first EOF mode for the case VAR (red), together with the 6-year forward-shifted and sign-inverted time series of normalized tidal mixing intensity (green).

[8] Strong tidal mixing makes isopycnal layers around 1026.8 kg m−3 thicker, and isopycnal surfaces above (below) that level become shallower (deeper). Such thickness anomaly is actually observed around the Kuril Islands and east coast of Japan [Yasuda et al., 2006; Osafune and Yasuda, 2006]. The anomaly propagates as coastal Kelvin waves, whose feature is clearly captured by additional experiments conducted by using only the ocean component of the coupled climate model (see Figure S1). The phase speed of their first baroclinic mode is of the order of 1 m s−1, so the signal reaches the equator within a year. Therefore, it is expected that anomaly of isopycnal surface depth, which should also be recognized as temperature anomaly, is found at the western end of the subsurface equatorial Pacific, and the strongest signal is realized around the timing of the strongest and weakest tidal mixing. Such variability is actually found in the case VAR. Statistically significant 18.6-year periodicity is found in temperature of the subsurface equatorial Pacific. Regression of annual-mean temperature to the 18.6-year tidal cycle (Figure 3; lag = 0 yr at equator) shows that the strongest signal is located at around 130°E and 150 m, the depth of the base of the warm pool. This means that the temperature at 150 m in this region becomes the lowest when the tidal mixing is strongest, and its amplitude (regression coefficient under the tidal amplitude normalized to unity) is ∼0.3°C.

Figure 3.

Lagged regression of annual-mean temperature in the upper equatorial Pacific along the equator and 2°N to the tidal cycle, whose amplitude is normalized to unity, with the tidal cycle leading. Colors indicate regression coefficient (in °C). The regions enclosed by white lines exhibit more than 90% significance. Black lines are for the long-term mean position of the isopycnal surfaces corresponding to 1025.0 and 1025.5 kg m−3.

[9] The temperature anomalies thus induced propagate eastward and upward along the equatorial thermocline (Figure 3; lag = 2, 4 yr at equator). Its relative slowness (e.g., 120° propagation in 5 years yields ∼0.3 m s−1) suggests that the propagation is due not to equatorial Kelvin waves but to advection. The anomalies are first transported northward by the New Guinea Coastal Current and then eastward by the northern part of Equatorial Under Current (EUC) along ∼2°N (Figure 3; lag = 0, 2, 4 at 2°N; see also Figure S2). The anomalies move equatorward due to subsurface equatorial convergence as they are transported eastward, and they finally reach the eastern equatorial subsurface. Since the anomalies are transported by the upper and northern part of EUC, where its speed is not very large (the modeled zonal velocity averaged along 2°N on the 1025.5 kg m−3 surface is ∼0.2 m s−1), it takes several years for the anomalies to propagate from the western end to the eastern end. Such advective control of anomaly propagation is confirmed by a tracer experiment (see Figure S2). Temperature anomaly eventually appears in the Pacific equatorial sea surface and transported westward (Figure 3; lag = 6 yr at equator) by the South Equatorial Current. At this stage, an anomaly of the opposite sign appears in the western subsurface, and the cycle goes into the opposite phase. As a whole, the effect of strong (weak) tides at the Kuril Islands appears as low (high) equatorial SST with about a 6-year lag. The modeled amplitude of 18.6-year periodicity in the eastern equatorial subsurface temperature is ∼0.1°C, and such magnitude is significant when decadal or longer climate variability is focused on.

4. Conclusion and Discussion

[10] The modeling result indicates that the 18.6-year cycle in strong tidal mixing localized around the Kuril Islands induces 18.6-year periodicity in ENSO-like Pacific climate variability. Influence of the tidal mixing first propagates along the Pacific western rim as coastally trapped waves. Temperature anomaly is generated in the subsurface western equatorial Pacific, which is then advected along the equatorial thermocline and eventually excites 18.6-year periodicity in the equatorial sea surface temperature.

[11] The governing physics involved herein is very simple and robust: coastal Kelvin waves and the EUC. These features are reasonably captured by most of state-of-the-art climate models, so are not very sensitive to model configurations. However, the EUC tends to be slow and broad in coarse-resolution ocean models as used in this study. A finer resolution climate model could result in a slightly different time lag between the tidal cycle and the phase of the 18.6-year period ENSO-like SST anomalies. Furthermore, much stronger signal might be found with a finer resolution climate model, as Kelvin waves tend to be diffuse in coarse-resolution ocean models.

[12] Lagged correlation between the time series of the first EOF mode of the observed SST (Figure 1a) and the tidal cycle peaks at a 4-year lag, which is slightly shorter than the current modeling result, but its statistical significance is low. Since reliable SST data are limited to recent 50–60 years, it is difficult at this stage to conclude whether or not tidally regulated bidecadal periodicity exists in the ENSO-like climate variability. An observation based analysis finds significant correlation between the winter PNA-index and the tidal cycle at a 2-year lag [McKinnell and Crawford, 2007]. The present modeling result also shows a correlation peak at a 2-year lag, but its statistical significance is again low. This study is only a first step to investigate the physical link between the 18.6-year period tides and bidecadal climate variability. Further studies, perhaps with improved models, are required to address the above issues. Furthermore, this study deals only with a one-way regulation mechanism from the ocean to the atmosphere. It is also necessary to investigate how air-sea coupling affects this regulation mechanism.

[13] Several air-sea coupling mechanisms have been proposed for decadal-scale Pacific climate variability through modeling [Wu et al., 2003]. This modeling study, for the first time, tests one-way regulation from the ocean to the climate through the combination of astronomical forcing and localized oceanic tidal mixing, which has been hypothesized by observation based studies [e.g., Yasuda et al., 2006; McKinnell and Crawford, 2007], and presents its possible mechanism. It does not oppose air-sea coupling mechanisms, however. The climate over the Pacific seems to intrinsically have ENSO-like variability of decadal to multidecadal time scales, as exemplified by the fact that even the case CONST exhibits such variability. The present one-way mechanism seems not to excite ENSO-like variability itself but to regulate periodicity of such intrinsic variability. Still, the view presented by this study would become a cornerstone of investigation into climate variability. The 18.6-year tides are predictable, so further studies along this line could shed new light on predictability of low-frequency climate variability.


[14] This work is supported by JST/CREST and MEXT/KAKENHI. The calculation herein is performed on NEC SX-6 at NIES. Development of MIROC was supported by MEXT/RR2002. We thank all the members of the MIROC development group and thank also S. Osafune and S. Minobe for comments and suggestions. Comments from two anonymous reviewers helped much to improve the manuscript and our understanding.